(A) Field of the Invention
This invention describes applications of monolithic 3D integration to various disciplines, including but not limited to, for example, light-emitting diodes, displays, image-sensors and solar cells.
(B) Discussion of Background Art
Semiconductor and optoelectronic devices often require thin monocrystalline (or single-crystal) films deposited on a certain wafer. To enable this deposition, many techniques, generally referred to as layer transfer technologies, have been developed. These include:    Ion-cut, variations of which are referred to as smart-cut, nano-cleave and smart-cleave: Further information on ion-cut technology is given in “Frontiers of silicon-on-insulator,” J. Appl. Phys. 93, 4955-4978 (2003) by G. K. Celler and S. Cristolovean (“Celler”) and also in “Mechanically induced Si layer transfer in hydrogen-implanted Si wafers,” Appl. Phys. Lett., vol. 76, pp. 2370-2372, 2000 by K. Henttinen, I. Suni, and S. S. Lau (“Hentinnen”).    Porous silicon approaches such as ELTRAN: These are described in “Eltran, Novel SOI Wafer Technology”, JSAP International, Number 4, July 2001 by T. Yonehara and K. Sakaguchi (“Yonehara”).    Lift-off with a temporary substrate, also referred to as epitaxial lift-off This is described in “Epitaxial lift-off and its applications”, 1993 Semicond. Sci. Technol. 8 1124 by P. Demeester, et al (“Demeester”).    Bonding a substrate with single crystal layers followed by Polishing, Time-controlled etch-back or Etch-stop layer controlled etch-back to thin the bonded substrate: These are described in U.S. Pat. No. 6,806,171 by A. Ulyashin and A. Usenko (“Ulyashin”) and “Enabling SOI-Based Assembly Technology for Three-Dimensional (3D) Integrated Circuits (ICs),” IEDM Tech. Digest, p. 363 (2005) by A. W. Topol, D. C. La Tulipe, L. Shi, S. M. Alam, D. J. Frank, S. E. Steen, J. Vichiconti, D. Posillico, M. Cobb, S. Medd, J. Patel, S. Goma, D. DiMilia, M. T. Robson, E. Duch, M. Farinelli, C. Wang, R. A. Conti, D. M. Canaperi, L. Deligianni, A. Kumar, K. T. Kwietniak, C. D'Emic, J. Ott, A. M. Young, K. W. Guarini, and M. Ieong (“Topol”).    Bonding a wafer with a Gallium Nitride film epitaxially grown on a sapphire substrate followed by laser lift-off for removing the transparent sapphire substrate: This method may be suitable for deposition of Gallium Nitride thin films, and is described in U.S. Pat. No. 6,071,795 by Nathan W. Cheung, Timothy D. Sands and William S. Wong (“Cheung”).    Rubber stamp layer transfer: This is described in “Solar cells sliced and diced”, 19 May 2010, Nature News.With novel applications of these methods and recognition of their individual strengths and weaknesses, one can significantly enhance today's light-emitting diode (LED), display, image-sensor and solar cell technologies.Background on LEDs
Light emitting diodes (LEDs) are used in many applications, including automotive lighting, incandescent bulb replacements, and as backlights for displays. Red LEDs are typically made on Gallium Arsenide (GaAs) substrates, and include quantum wells constructed of various materials such as AlInGaP and GaInP. Blue and green LEDs are typically made on Sapphire or Silicon Carbide (SiC) or bulk Gallium Nitride (GaN) substrates, and include quantum wells constructed of various materials such as GaN and InGaN.
A white LED for lighting and display applications can be constructed by either using a blue LED coated with phosphor (called phosphor-coated LED or pcLED) or by combining light from red, blue, and green LEDs (called RGB LED). RGB LEDs are typically constructed by placing red, blue, and green LEDs side-by-side. While RGB LEDs are more energy-efficient than pcLEDs, they are less efficient in mixing red, blue and green colors to form white light. They also are much more costly than pcLEDs. To tackle issues with RGB LEDs, several proposals have been made.
One RGB LED proposal from Hong Kong University is described in “Design of vertically stacked polychromatic light emitting diodes”, Optics Express, June 2009 by K. Hui, X. Wang, et al (“Hui”). It involves stacking red, blue, and green LEDs on top of each other after individually packaging each of these LEDs. While this solves light mixing problems, this RGB-LED is still much more costly than a pcLED solution since three LEDs for red, blue, and green color need to be packaged. A pcLED, on the other hand, requires just one LED to be packaged and coated with phosphor.
Another RGB LED proposal from Nichia Corporation is described in “Phosphor Free High-Luminous-Efficiency White Light-Emitting Diodes Composed of InGaN Multi-Quantum Well”, Japanese Journal of Applied Physics, 2002 by M. Yamada, Y. Narukawa, et al. (“Yamada”). It involves constructing and stacking red, blue and green LEDs of GaN-based materials on a sapphire or SiC substrate. However, red LEDs are not efficient when constructed with GaN-based material systems, and that hampers usefulness of this implementation. It is not possible to deposit defect-free AlInGaP/InGaP for red LEDs on the same substrate as GaN based blue and green LEDs, due to a mismatch in thermal expansion co-efficient between the various material systems.
Yet another RGB-LED proposal is described in “Cascade Single chip phosphor-free while light emitting diodes”, Applied Physics Letters, 2008 by X. Guo, G. Shen, et al. (“Guo”). It involves bonding GaAs based red LEDs with GaN based blue-green LEDs to produce white light. Unfortunately, this bonding process requires 600° C. temperatures, causing issues with mismatch of thermal expansion co-efficients and cracking Another publication on this topic is “A trichromatic phosphor-free white light-emitting diode by using adhesive bonding scheme”, Proc. SPIE, Vol. 7635, 2009 by D. Chuai, X. Guo, et al. (“Chuai”). It involves bonding red LEDs with green-blue LED stacks. Bonding is done at the die level after dicing, which is more costly than a wafer-based approach.
U.S. patent application Ser. No. 12/130,824 describes various stacked RGB LED devices. It also briefly mentions a method for construction of a stacked LED where all layers of the stacked LED are transferred using lift-off with a temporary carrier and Indium Tin Oxide (ITO) to semiconductor bonding. This method has several issues for constructing a RGB LED stack. First, it is difficult to manufacture a lift-off with a temporary carrier of red LEDs for producing a RGB LED stack, especially for substrates larger than 2 inch. This is because red LEDs are typically constructed on non-transparent GaAs substrates, and lift-off with a temporary carrier is done by using an epitaxial lift-off process. Here, the thin film to be transferred typically sits atop a “release-layer” (eg. AlAs), this release layer is removed by etch procedures after the thin film is attached to a temporary substrate. Scaling this process to 4 inch wafers and bigger is difficult. Second, it is very difficult to perform the bonding of ITO to semiconductor materials of a LED layer at reasonable temperatures, as described in the patent application Ser. No. 12/130,824.
It is therefore clear that a better method for constructing RGB LEDs will be helpful. Since RGB LEDs are significantly more efficient than pcLEDs, they can be used as replacements of today's phosphor-based LEDs for many applications, provided a cheap and effective method of constructing RGB LEDs can be invented.
Background on Image-Sensors:
Image sensors are used in applications such as cameras. Red, blue, and green components of the incident light are sensed and stored in digital format. CMOS image sensors typically contain a photodetector and sensing circuitry. Almost all image sensors today have both the photodetector and sensing circuitry on the same chip. Since the area consumed by the sensing circuits is high, the photodetector cannot see the entire incident light, and image capture is not as efficient.
To tackle this problem, several researchers have proposed building the photodetectors and the sensing circuitry on separate chips and stacking them on top of each other. A publication that describes this method is “Megapixel CMOS image sensor fabricated in three-dimensional integrated circuit technology”, Intl. Solid State Circuits Conference 2005 by Suntharalingam, V., Berger, R., et al. (“Suntharalingam”). These proposals use through-silicon via (TSV) technology where alignment is done in conjunction with bonding. However, pixel size is reaching the 1 μm range, and successfully processing TSVs in the 1 μm range or below is very difficult. This is due to alignment issues while bonding. For example, the International Technology Roadmap for Semiconductors (ITRS) suggests that the 2-4 um TSV pitch will be the industry standard until 2012. A 2-4 μm pitch TSV will be too big for a sub-1 μm pixel. Therefore, novel techniques of stacking photodetectors and sensing circuitry are required.
A possible solution to this problem is given in “Setting up 3D Sequential Integration for Back-Illuminated CMOS Image Sensors with Highly Miniaturized Pixels with Low Temperature Fully-depleted SOI Transistors,” IEDM, p. 1-4 (2008) by P. Coudrain et al. (“Coudrain”). In the publication, transistors are monolithically integrated on top of photodetectors. Unfortunately, transistor process temperatures reach 600° C. or more. This is not ideal for transistors (that require a higher thermal budget) and photodetectors (that may prefer a lower thermal budget).
Background on Displays:
Liquid Crystal Displays (LCDs) can be classified into two types based on manufacturing technology utilized: (1) Large-size displays that are made of amorphous/polycrystalline silicon thin-film-transistors (TFTs), and (2) Microdisplays that utilize single-crystal silicon transistors. Microdisplays are typically used where very high resolution is needed, such as camera/camcorder view-finders, projectors and wearable computers.
Microdisplays are made in semiconductor fabs with 200 mm or 300 mm wafers. They are typically constructed with LCOS (Liquid-Crystal-on-Silicon) Technology and are reflective in nature. An exception to this trend of reflective microdisplays is technology from Kopin Corporation (U.S. Pat. No. 5,317,236, filed December 1991). This company utilizes transmittive displays with a lift-off layer transfer scheme. Transmittive displays may be generally preferred for various applications.
While lift-off layer transfer schemes are viable for transmittive displays, they are frequently not used for semiconductor manufacturing due to yield issues. Therefore, other layer transfer schemes will be helpful. However, it is not easy to utilize other layer transfer schemes for making transistors in microdisplays. For example, application of “smart-cut” layer transfer to attach monocrystalline silicon transistors to glass is described in “Integration of Single Crystal Si TFTs and Circuits on a Large Glass Substrate”, IEDM 2009 by Y. Takafuji, Y. Fukushima, K. Tomiyasu, et al. (“Takafuji”). Unfortunately, hydrogen is implanted through the gate oxide of transferred transistors in the process, and this degrades performance. Process temperatures are as high as 600° C. in this paper, and this requires costly glass substrates. Several challenges therefore need to be overcome for efficient layer transfer, and require innovation.
Background on Solar Cells:
Solar cells can be constructed of several materials such as, for example, silicon and compound semiconductors. The highest efficiency solar cells are typically multi junction solar cells that are constructed of compound semiconductor materials. These multi junction solar cells are typically constructed on a germanium substrate, and semiconductors with various band-gaps are epitaxially grown atop this substrate to capture different portions of the solar spectrum.
There are a few issues with standard multi junction solar cells. Since multiple junctions are grown epitaxially above a single substrate (such as Germanium) at high temperature, materials used for different junctions are restricted to those that have lattice constants and thermal expansion co-efficients close to those of the substrate. Therefore, the choice of materials used to build junctions for multi-junction solar cells is limited. As a result, most multi junction solar cells commercially available today cannot capture the full solar spectrum. Efficiency of the solar cell can be improved if a large band of the solar spectrum is captured. Furthermore, multi junction solar cells today suffer from high cost of the substrate above which multiple junctions are epitaxially grown. Methods to build multi-junction solar cells that tackle both these issues will be helpful.
A method of making multi junction solar cells by mechanically bonding two solar cells, one with a Germanium junction and another with a compound semiconductor junction is described in “Towards highly efficient 4-terminal mechanical photovoltaic stacks”, III-Vs Review, Volume 19, Issue 7, September-October 2006 by Giovanni Flamand, Jef Poortmans (“Flamand”). In this work, the authors make the compound semiconductor junctions on a Germanium substrate epitaxially. They then etch away the entire Germanium substrate after bonding to the other substrate with the Germanium junction. The process uses two Germanium substrates, and is therefore expensive.
Techniques to create multi junction solar cells with layer transfer have been described in “Wafer bonding and layer transfer processes for 4-junction high efficiency solar cells,” Photovoltaic Specialists Conference, 2002. Conference Record of the Twenty-Ninth IEEE, vol., no., pp. 1039-1042, 19-24 May 2002 by Zahler, J. M.; Fontcuberta i Morral, A.; Chang-Geun Ahn; Atwater, H. A.; Wanlass, M. W.; Chu, C. and Iles, P. A. An anneal is used for ion-cut purposes, and this anneal is typically done at temperatures higher than 350-400° C. (if high bond strength is desired). When that happens, cracking and defects can be produced due to mismatch of co-efficients of thermal expansion between various layers in the stack. Furthermore, semiconductor layers are bonded together, and the quality of this bond not as good as oxide-to-oxide bonding, especially for lower process temperatures.